摘要： 为提高结合共享存储器和片上网络进行核间通信的多核处理器的通信带宽使用效率,基于异构通信机制,提出一种改进的任务映射算法。使用分支定界算法缩小搜索空间,通过迭代交换任务的位置提高映射质量。实验结果表明,与NMAP算法相比,该算法对于随机任务图的通信代价平均降低18.3%,对于并行排序应用最高降低30.6%。

关键词: 片上网络, 共享存储器, 任务映射算法, 异构通信机制, 多核处理器

Abstract: In order to take full advantage of the communication bandwidth,this paper presents an improved task mapping algorithm based on heterogeneous communication mechanisms for a multicore processor with both Network on Chip(NoC) and shared-memory inter-core communication mechanisms.The proposed algorithm uses the branch and bound algorithm to reduce the search space,and then improves the quality of mapping by iteratively exchanging the positions of the tasks.Experimental results show that,compared with the NMAP algorithm,the communication cost of the proposed algorithm is reduced by an average of 18.3% for random task graphs,and the communication cost can be reduced by up to 30.6% for parallel sorting applications.

Key words: Network on Chip(NoC), shared memory, task mapping algorithm, heterogeneous communication mechanism, multicore processor

中图分类号: 

• TN911